Why Frontier AI Labs Are Defying the Open Source Threat

The data from Vercel’s AI gateway suggests a clear divide between usage and revenue. While models like DeepSeek now dominate token volumes—processing over a third of traffic on the platform—Anthropic still commands more than half of total AI spending. This disparity persists because frontier models remain significantly more expensive; OpenRouter data shows that the Opus 4.8 model costs roughly 23 times more per million tokens than DeepSeek’s V4 Flash.

This economic resilience stems from the rapid expansion of AI-addressable tasks. As developers find new, high-stakes use cases, they gravitate toward the performance of frontier models for initial development. Even when mature workflows migrate to lighter, open source tools, the constant influx of new, difficult problems keeps the top labs at the center of enterprise budgets. Rather than being reduced to commodity suppliers, companies like Anthropic appear to be successfully maintaining their position as the premium tier of the market, effectively owning the discovery phase of the AI lifecycle.